2017 12 06

wangkuiyi

• iOS build document: https://github.com/PaddlePaddle/Paddle/pull/6180
• Documentation
  • https://github.com/PaddlePaddle/Paddle/pull/6306#pullrequestreview-81360817
  • https://github.com/PaddlePaddle/Paddle/issues/6311

caoying

• PR review

  • Review and help to refine the doc of "TensorFlow model to Paddle model":
    • https://github.com/PaddlePaddle/models/pull/483#pullrequestreview-80742107
  • The FM layer:
    • https://github.com/PaddlePaddle/Paddle/pull/4859#pullrequestreview-79106694
  • DeepFM for CTR:
    • https://github.com/PaddlePaddle/models/pull/485#pullrequestreview-79100390
  • math function for sampling integers：
    • https://github.com/PaddlePaddle/Paddle/pull/5945#pullrequestreview-80418358

• WIP: the documentation of C-API

  • https://github.com/PaddlePaddle/Paddle/compare/develop...lcy-seso:update_capi_doc

• Proofread book chapter 7 & 8 for a book written by colleagues from the department of engineering efficiency (gong cheng xiao lv bu).

sweetsky0901(tiantian)

• spp(Spatial pyramid pooling) op, doing, code review https://github.com/PaddlePaddle/Paddle/issues/5988
• detection_output op（for SSD, doing, code 35% https://github.com/PaddlePaddle/Paddle/issues/6225

Yu Yang

• Use WhileOp to create dynamic RNN
  • https://github.com/PaddlePaddle/Paddle/pull/6140
  • https://github.com/PaddlePaddle/Paddle/pull/6282
• Fix performance problem in PADDNE_ENFORCE
  • https://github.com/PaddlePaddle/Paddle/pull/6085
• Expose some operator to layers
  • Also polish API consistency
  • https://github.com/PaddlePaddle/Paddle/pull/6147
• Add DataFeeder
  • https://github.com/PaddlePaddle/Paddle/pull/6102
• Serveral Enhancements
  • https://github.com/PaddlePaddle/Paddle/pull/6343
  • https://github.com/PaddlePaddle/Paddle/pull/6112
  • https://github.com/PaddlePaddle/Paddle/pull/6084

ranqiu

• Proofread book chapter 5 & 6 for a book written by colleagues from the department of engineering efficiency (gong cheng xiao lv bu).

• Tune understand_sentiment on fluid to catch up with the performance on PaddlePaddle V2

Yancey1989(Yan Xu)

• Muti-CPU design doc, https://github.com/PaddlePaddle/Paddle/pull/6223
• Add an Android project on TeamCity, https://github.com/PaddlePaddle/Paddle/issues/6111
• Add version interface in Python API, https://github.com/PaddlePaddle/Paddle/pull/2985
• BugFix:
  • fix gprc compile warning, https://github.com/PaddlePaddle/Paddle/pull/6050
  • [Review]Fix can not find cblas library,https://github.com/PaddlePaddle/Paddle/pull/6347
• Review:
  • grpc benchmark, https://github.com/PaddlePaddle/Paddle/pull/6176

gongweibao

• Add LRN efficient GPU implement.
  • https://github.com/PaddlePaddle/Paddle/pull/5894
• PaddleCloud
  • Fix ending slash bugs: https://github.com/PaddlePaddle/cloud/pull/513
• Cluster:
  • Deploy Dashboard on dlnel cluster.
  • Deploy Ingress controller and pfs server on Mulan-CPU cluster.

luotao

• add WITH_DOC for print_operators_doc: https://github.com/PaddlePaddle/Paddle/pull/6184
• simplify the CMakeLists.txt of trainer/tests and gserver/tests: https://github.com/PaddlePaddle/Paddle/pull/5830
• fix typo in pip_install_cn/en.rst: https://github.com/PaddlePaddle/Paddle/pull/6263
• code review:
  • Proofread book chapter 9 & 10 for a book written by colleagues from the department of engineering efficiency (gong cheng xiao lv bu).
  • mkl:
    • use awk command to replace bc: https://github.com/PaddlePaddle/Paddle/pull/6222
    • add script to check the cpu env: https://github.com/PaddlePaddle/Paddle/pull/6291
  • doc:
    • Update write docs cn: https://github.com/PaddlePaddle/Paddle/pull/6014
    • Generate docs for Fluid API: https://github.com/PaddlePaddle/Paddle/pull/6215

wanghaoshuang

• Add a tutorial for policy gradient
  • https://github.com/PaddlePaddle/models/pull/505
• Make ConvTransProjection support for dilation
  • https://github.com/PaddlePaddle/Paddle/pull/6188
• Fix nce op
  • https://github.com/PaddlePaddle/Paddle/pull/6333

qijun

Multi-device design:

• python demo code for setting multi-device
  • https://github.com/PaddlePaddle/Paddle/pull/6143
• Discussion on ExecutionPlan

Fluid docs:

• Generate docs for Fluid API
  • https://github.com/PaddlePaddle/Paddle/pull/6215
• build paddle_python before build paddle_docs
  • https://github.com/PaddlePaddle/Paddle/pull/6337

GPU memory allocation:

• change GPU memory allocating policy
  • https://github.com/PaddlePaddle/Paddle/pull/6159
• fix bug in default GPU memory allocating policy
  • https://github.com/PaddlePaddle/Paddle/pull/6268

Fix and Enhance:

• open test_word2vec
  • https://github.com/PaddlePaddle/Paddle/pull/6104
• disable test_recurrent_op unittest
  • https://github.com/PaddlePaddle/Paddle/pull/6153

typhoonzero(wuyi):

• discussions about multi thread and distributed:
  • https://github.com/PaddlePaddle/Paddle/projects/30
• API for distributed training, WIP:
  • https://github.com/PaddlePaddle/Paddle/pull/6146
  • benchmark: https://github.com/PaddlePaddle/Paddle/issues/6176
  • https://github.com/PaddlePaddle/Paddle/pull/6297
• Evaluating DeepFM using feed ctr data on cluster

yangyaming

• Doc unification for fluid ops
  https://github.com/PaddlePaddle/Paddle/issues/6160
  https://github.com/PaddlePaddle/Paddle/issues/6361
• Resource releasing problem after process exit
  https://github.com/PaddlePaddle/DeepSpeech/pull/58
• Chapters convergence validationcode review
  https://github.com/PaddlePaddle/Paddle/issues/6208
• Code Review
  https://github.com/PaddlePaddle/DeepSpeech/pull/52#pullrequestreview-80744617
  https://github.com/dzhwinter/benchmark/pull/5#pullrequestreview-80747588
  https://github.com/PaddlePaddle/DeepSpeech/pull/51#pullrequestreview-80795581
  https://github.com/dzhwinter/benchmark/pull/8#pullrequestreview-81205628
  https://github.com/dzhwinter/benchmark/pull/9#pullrequestreview-81492918
  https://github.com/PaddlePaddle/Paddle/pull/6150#pullrequestreview-81406649

guosheng

• PR:
  • Enhance AvgPooling in PoolLayer to support include_mode(Merged)
    • https://github.com/PaddlePaddle/Paddle/pull/6100
  • Add the Inception-ResNet-v2 model
    • https://github.com/PaddlePaddle/models/pull/510
  • Add ChunkEvaluator for multi-batch(WIP)
• PR Review:
  • https://github.com/PaddlePaddle/Paddle/pull/6296
  • https://github.com/PaddlePaddle/Paddle/pull/6198
• Others:
  • Run SNPE demo for Caffe Alexnet

Zhangchao

• PR

  • Fix shared parameters in nested text classification.

    https://github.com/PaddlePaddle/models/pull/513

  • Fix the default model dir in NER.

    https://github.com/PaddlePaddle/models/pull/517

• Proofread book chapter 3 & 4 for a book written by colleagues from the department of engineering efficiency (gong cheng xiao lv bu).

• Valid performance of word2vec between Fluid and PaddlePaddle V2.

Chenxi

• ExecutionPlan Discussion

  • https://github.com/PaddlePaddle/Paddle/pull/6078

• Minior fix on doc

  • https://github.com/PaddlePaddle/Paddle/pull/6318

• Imperative Programming doc refinement

  • https://github.com/PaddlePaddle/Paddle/pull/6076

Dang Qingqing

• Fix the doc of LSTM operator: https://github.com/PaddlePaddle/Paddle/pull/6148
• Make lstm_op follow google code style: https://github.com/PaddlePaddle/Paddle/pull/6198
• Make the new framework independent the old framework. https://github.com/PaddlePaddle/Paddle/pull/6201
• Work about benchmark:
  • Refine and speedup momentum operator. https://github.com/PaddlePaddle/Paddle/pull/6233
  • Remove the CUDA stream synchronization between each operator. https://github.com/PaddlePaddle/Paddle/pull/6284
  • Change the type of conv2d in Python API. https://github.com/PaddlePaddle/Paddle/pull/6336
  • Fluid and Tensorflow comparison for CNN model with Longfei: https://github.com/PaddlePaddle/Paddle/issues/5892
  • some code review: https://github.com/dzhwinter/benchmark/pulls?q=is%3Apr+is%3Aclosed
  • Port the timer tools of old PaddlePaddle framework into Fluid: https://github.com/qingqing01/Paddle/tree/add_timer But Yibing and I will implement more flexiable profling tools for fluid [doing].
  • C++ and Python profling: https://github.com/PaddlePaddle/Paddle/issues/6179
• Work about mobile
  • Use protobuf v3.2.0 for MOBILE_INFERENCE compiling. https://github.com/PaddlePaddle/Paddle/pull/6115
  • Release C-API for android with Yiqun and Daoyuan: http://icode.baidu.com/repos/baidu/paddle/mobile/tree/master
• Code Review:
  • Add row conv operator : https://github.com/PaddlePaddle/Paddle/pull/6013
  • Add LRN efficient GPU implement: https://github.com/PaddlePaddle/Paddle/pull/5894
  • fix bug in default GPU memory allocating policy: https://github.com/PaddlePaddle/Paddle/pull/6268
  • other code.

dongzhihong

• Fluid
  • add model average layer
    • https://github.com/PaddlePaddle/Paddle/pull/6365
  • add multi kernel register
    • https://github.com/PaddlePaddle/Paddle/pull/6192
  • rewrited Python datatype
    • https://github.com/PaddlePaddle/Paddle/pull/6117
  • add program level init seed
    • https://github.com/PaddlePaddle/Paddle/pull/6221
  • fix build cares
    • https://github.com/PaddlePaddle/Paddle/pull/6097
• Profiler/Chapters Performance

  • Chapters Performance

    • https://github.com/PaddlePaddle/Paddle/issues/6208

  • benchmark repos

    • https://github.com/dzhwinter/benchmark

  • add tensorflow lstm

    • https://github.com/dzhwinter/benchmark/pull/4

  • add paddle dynamic lstm /static lstm

    • https://github.com/dzhwinter/benchmark/pull/3
    • https://github.com/dzhwinter/benchmark/pull/2

  • fixes

    • https://github.com/dzhwinter/benchmark/pull/8
    • https://github.com/dzhwinter/benchmark/pull/9
    • https://github.com/dzhwinter/benchmark/pull/5

  • Multi-GPU/Distributed Training

    • https://github.com/PaddlePaddle/Paddle/issues/6319

wangmeng

• Reorganize the ltr model:
  • https://github.com/PaddlePaddle/models/pull/515
• Revise the textbook on deep learning & paddle written by department of engineering efficiency and beihang
• Revise the code corresponding to the textbook:
  • https://github.com/BaiduOSS/DeepLearningAndPaddleTutorial/pull/5
  • https://github.com/BaiduOSS/DeepLearningAndPaddleTutorial/pull/4

hedaoyuan

• Mobile
  Promote the reconstruction of Paddle inference interfaces. With more and more users using Paddle inference, inference interfaces expose a number of issues that require clearer interface descriptions and documentation to reduce user-related issues.
• Review
  • https://github.com/PaddlePaddle/Paddle/pull/6340
  • https://github.com/PaddlePaddle/Paddle/pull/6177
  • https://github.com/PaddlePaddle/Paddle/pull/5716

helinwang

• ExecutionPlan design doc PR discussion: https://github.com/PaddlePaddle/Paddle/pull/6078
• Compile time runtime separation discussions: https://github.com/PaddlePaddle/Paddle/issues/6359

Yibing Liu

Deep Speech 2:

• Correct the error rate's computation for multiple sentences
  • https://github.com/PaddlePaddle/DeepSpeech/pull/51
• Add infer & test scripts for baidu_en8k model
  • https://github.com/PaddlePaddle/DeepSpeech/pull/52
• Retune hyper-parameters and update benchmark results for English models
  • https://github.com/PaddlePaddle/DeepSpeech/pull/55
• Code Review
  • https://github.com/PaddlePaddle/DeepSpeech/pull/47
  • https://github.com/PaddlePaddle/DeepSpeech/pull/58

Paddle:

• Performance test of GoogleNet, ResNet，VGG on Mac with mkl_dnn
  • https://github.com/PaddlePaddle/Paddle/issues/6202
• Performance tuning of fluid

xingzhaolong

• Enet:

  • prelu neon pr https://github.com/PaddlePaddle/Paddle/pull/6177
  • Adjust the model to apply to the mobile
  • Testing enet in multiple compiling environments
    • armv7 41.671
    • armv8 29.7395
    • nnpack-v7 48.6846
    • nnpack-v8 31.4138

• add pascal ssd mobilenet model https://github.com/PaddlePaddle/Mobile/pull/40

zhaochengduo

Profiler Performance

• Refine enforce.h
  • https://github.com/PaddlePaddle/Paddle/pull/6164
• Update elementwise op
  • https://github.com/PaddlePaddle/Paddle/pull/6229
• Refine code（operator++）
  • https://github.com/PaddlePaddle/Paddle/pull/6324

Summary

• A summary of issues about C-API
  • https://github.com/PaddlePaddle/Paddle/wiki/A-summary-of-issues-about-CAPI

Enhancement

• Add dilation for conv_trans_op [WIP]
  • https://github.com/PaddlePaddle/Paddle/pull/6279

Review

• Refine and speedup momentum operator.
  • https://github.com/PaddlePaddle/Paddle/pull/6233
• Add benchmark configuration and scripts for convnets.
  • https://github.com/PaddlePaddle/Paddle/pull/6056

Liu Yiqun

• Refactor
  • [Reviewing] Port WarpCTC Operator
    • https://github.com/PaddlePaddle/Paddle/pull/5107
  • Review
    • Compile cblas library as static: https://github.com/PaddlePaddle/Paddle/pull/6347
    • Add ctc edit distance operator: https://github.com/PaddlePaddle/Paddle/pull/5300
• Mobile deployment
  • Add the support of cross-compiling for Android using Android NDK directly
  • Compiling multi-thread openblas for AR: 50ms (single-thread) vs 35ms (multi-thread openblas)
  • Review
    • Init Commit for iOS AICamera Demo: https://github.com/PaddlePaddle/Mobile/pull/26
    • Correct the README of AICamera: https://github.com/PaddlePaddle/Mobile/pull/44
    • Translate the iOS build doc: https://github.com/PaddlePaddle/Paddle/pull/6180
    • Update write docs cn: https://github.com/PaddlePaddle/Paddle/pull/6014

qiaolongfei

refactoring

• feature

  • add crf_decoding layer
  • add chunk eval layer
  • fix int overflow
  • Nmt decoder train

• profile

  • prepare a machine with docker and GPU for profiling
  • image classification benchmark and finetune
  • Benchmark convnet model tf

• multi GPU/machine training

  • some discussion and review

• review

  • Add ExecutionPlan design
  • Change the type of conv2d in Python API
  • feature/nmt add encoder
  • Feature/while op sentiment analysis
  • Feature/simple gan for api

Yan Chunwei

• WIP model compression survey

• chapter 8 chapter translation status, decoder config finished, waiting for encoder valid

• PR, add gru unit layer wrapper

• PR, chapter 8 machine translation, encoder

• PR, visualDL, add write read support

• discussion, modelcontiguous integration framework

abhinavarora

• Documentation
  • Polishing the cpu profiling doc https://github.com/PaddlePaddle/Paddle/pull/6116
  • Polish the Evaluator design doc https://github.com/PaddlePaddle/Paddle/pull/6195
  • Using latex fraction syntax in sigmoid and logsigmoid op https://github.com/PaddlePaddle/Paddle/pull/6292
  • Fix the rendering of latex equation for adamax op https://github.com/PaddlePaddle/Paddle/pull/6294
  • Fix equation in logical or op https://github.com/PaddlePaddle/Paddle/pull/6315
  • Fix equations in sequence_pool op https://github.com/PaddlePaddle/Paddle/pull/6355
• Profiling
  • Changing RelWithDebInfo flags https://github.com/PaddlePaddle/Paddle/pull/6193
• Operators
  • Swish Activation op https://github.com/PaddlePaddle/Paddle/pull/6358
• PR Review
  • https://github.com/PaddlePaddle/Paddle/pull/6073
  • https://github.com/PaddlePaddle/Paddle/pull/6215
  • https://github.com/PaddlePaddle/Paddle/pull/6302
  • https://github.com/PaddlePaddle/Paddle/pull/6303
  • https://github.com/PaddlePaddle/Paddle/pull/6304
  • https://github.com/PaddlePaddle/Paddle/pull/6307
  • https://github.com/PaddlePaddle/Paddle/pull/6312
  • https://github.com/PaddlePaddle/Paddle/pull/6314

cs2be (thuan)

• Expose "Build using Docker" documentation on PaddlePaddle.org https://github.com/PaddlePaddle/Paddle/pull/6306
• Implement crude PaddlePaddle Program to Tensorflow graph convertor. https://github.com/PaddlePaddle/board/pull/14